Publication number | US20060007990 A1 |

Publication type | Application |

Application number | US 11/219,626 |

Publication date | Jan 12, 2006 |

Filing date | Sep 2, 2005 |

Priority date | Mar 12, 2004 |

Also published as | EP1886413A1, US7848389, WO2006128490A1 |

Publication number | 11219626, 219626, US 2006/0007990 A1, US 2006/007990 A1, US 20060007990 A1, US 20060007990A1, US 2006007990 A1, US 2006007990A1, US-A1-20060007990, US-A1-2006007990, US2006/0007990A1, US2006/007990A1, US20060007990 A1, US20060007990A1, US2006007990 A1, US2006007990A1 |

Inventors | Carmela Cozzo, Douglas Cairns, Gregory Bottomley, Ali Khayrallah, Hakan Eriksson |

Original Assignee | Telefonaktiebolaget Lm |

Export Citation | BiBTeX, EndNote, RefMan |

Patent Citations (24), Referenced by (39), Classifications (12), Legal Events (2) | |

External Links: USPTO, USPTO Assignment, Espacenet | |

US 20060007990 A1

Abstract

A wireless communication receiver, such as the receiver included in a wireless communication transceiver implemented in a base station or in a mobile station of a wireless communication network, includes a parametric G-RAKE receiver circuit and a method that compute parametric scaling parameters on a per transmission interval basis. In one embodiment, measured impairment correlations are obtained for an individual transmission slot and used to estimate instantaneous values of the scaling parameters. One or both of those instantaneous values are then constrained according to one or more defined limits. In other embodiments, multiple transmission slots are used to increase the number of measurements available to estimate the scaling parameters, with parameter constraining optionally applied. Further embodiments use iterative methods and/or solve for one parameter, and use the results to obtain the other parameter(s). One or more of these embodiments can be improved through the use of error correction/detection information.

Claims(33)

for each of one or more transmission intervals of a received communication signal, obtaining measured impairment correlations over all or part of the transmission interval;

estimating the scaling parameter by expressing the measured impairment correlations as a function of modeled impairment correlations as scaled by an unknown instantaneous value of the scaling parameter and solving for the instantaneous value; and

bounding the instantaneous value to keep it within one or more desired ranges.

Description

The instant application is a continuation-in-part of and claims priority under 35 U.S.C. § 120 from the earlier-filed U.S. patent application entitled “Method and Apparatus for Parameter Estimation in a Generalized RAKE Receiver,” which was filed on 12 Mar. 2004, and is assigned application Ser. No. 10/800,167. The instant application further claims priority under 35 U.S.C. 119(e) from the provisional patent application entitled “Estimation of Scaling Parameters for Parametric G-RAKE Receiver,” which was filed on 31 May 2005 and is assigned application Ser. No. 60/685,825. These related applications are incorporated in their entireties by reference herein.

The present invention relates to wireless communication receivers, and particularly relates to scaling parameter estimation in Generalized RAKE receivers.

With multipath propagation, a receiver receives multiple images of the transmitted signal at different delays corresponding to the different propagation paths. “Standard” RAKE receivers improve reception performance in multipath environments by time aligning each of one or more RAKE fingers with a corresponding one of the propagation path delays. Each RAKE finger outputs despread traffic values from the signal image corresponding to that propagation path delay, and a combining circuit forms a combined signal based on combining the finger outputs according to a set of combining weights. The combining weight assigned to each finger generally is computed as the complex conjugate of the channel tap calculated for that finger's delay path.

If the received signal impairments are uncorrelated across the RAKE fingers, the standard RAKE receiver solution is optimal. However, the presence of correlated impairments in the received signal degrades the performance of the standard RAKE receiver solution, and the performance degradation can be substantial.

For various reasons, such as the use of higher transmission bandwidths for increased data rates, receivers intended for use in the evolving wireless communication networks are more likely to “see” significant levels of colored noise and other correlated impairments in their received signals. As such, the standard RAKE receiver often is not a suitable candidate architecture for receivers intended to operate in such circumstances.

A newer approach, often termed a “Generalized” RAKE (G-RAKE) receiver, improves reception performance in dispersive channel environments by offering a combination of interference cancellation and channel equalization. To gain these improvements over the standard RAKE receiver architecture, the G-RAKE receiver modifies its signal processing operations in a number of significant ways. First, one or more of its fingers are placed off-path, i.e., offset from the path delays measured for the received signal. Second, its combining weights are not strictly channel tap conjugates. Rather, the combining weights are based at least in part on impairment correlation estimates, which allow the combining process to reduce correlated interference in the combined signal.

One type of G-RAKE recognizes that the impairment correlations can be decomposed into constituent elements, and that the underlying “structure” of these constituent elements can be modeled. As used herein, the term “impairment” has a broad definition that includes, but is not limited to, one or more of the following items: adjacent system interference, self and multi-user interference, and noise. Because modeled terms are used in their impairment correlation estimations, these types of G-RAKE receivers are broadly referred to as “parametric” G-RAKE receivers. As an example, the impairment correlations measured for the received signal may be expressed as the sum of modeled interference correlations as scaled by a corresponding model fitting parameter and modeled noise correlations as scaled by a corresponding model fitting parameter. Because the structures of the modeled terms are known, and the short-term impairment correlations may be measured from a set of despread pilot values, for example, the impairment correlation estimation task is reduced to the determination of the appropriate model fitting parameters, which are also referred to as scaling parameters.

However, challenges remain in parametric G-RAKE processing. More particularly, certain conditions make it difficult to maintain appropriately updated scaling parameters. In particular, the scaling parameters depend on the total energy per chip period of the signal, E_{c}, and on the white noise or simply noise power (thermal noise plus other interference), N_{0}. These values change rapidly under certain circumstances, making it difficult for parametric G-RAKE receivers to maintain appropriately updated scaling parameters. Examples include the varying transmission conditions, e.g., widely varying transmission powers, associated with scheduled and non-scheduled users on high-rate shared data channels, such as the High Speed Downlink Shared Channel (HS-DSCH) for High Speed Downlink Packet Access (HSDPA) introduced in Release 5 of the Wideband CDMA (WCDMA) standards. Further, the introduction of the Enhanced Up-Link (EUL) in Release 6 of the WCDMA standards, which also includes strict user scheduling, portends complications for parameter estimation in G-RAKE receiver applications.

In at least one embodiment of a parametric Generalized RAKE (G-RAKE) receiver circuit as taught herein, parametric scaling parameters used to scale the modeled impairment correlations that, in turn, are used to generate impairment correlation estimates for combining weight computation, are estimated on a per transmission interval basis. As such, the potentially dramatic interval-to-interval changes in the signal conditions bearing on the calculation of the scaling parameters do not upset the scaling parameter estimation process.

Thus, in one embodiment, one or more processing circuits comprising a parametric G-RAKE receiver circuit are configured to estimate the scaling parameter(s) on a per transmission interval basis. Doing so prevents the scaling parameters estimated in one transmission interval of the received communication signal from being influenced by the scaling parameters estimated in another transmission interval of the received communication signal.

In another embodiment, a method of estimating a scaling parameter used to scale modeled impairment correlations in a parametric Generalized RAKE (G-RAKE) receiver comprises, for each of one or more transmission intervals of a received communication signal, obtaining measured impairment correlations over all or part of the transmission interval. The method continues with estimating the scaling parameter by expressing the measured impairment correlations as a function of modeled impairment correlations as scaled by an unknown instantaneous value of the scaling parameter and solving for the instantaneous value.

For example, the measured impairment correlations may be expressed as a function of modeled interference correlations scaled by a first scaling parameter and modeled noise correlations scaled by a second scaling parameter. Instantaneous values for the two scaling parameters may be determined by performing a Least Squares (LS) fitting process. In one variation, one parameter is expressed in terms of the other, and the LS fitting process reduces to solving for one scaling parameter. In other variations, iterative approaches may be used to solve for the parameter(s).

Further, error correction/detection information in the received communication signal, such as Cyclic Redundancy Check (CRC) values, may be used to select candidate scaling parameters from a set of candidates. For example, candidate scaling parameter values can be used to generate combining weights, and the CRCs in decoded information obtained from the received symbol estimates generated from those combining weights can be checked.

Of course, the present invention is not limited by the above features and advantages. Indeed, those skilled in the art will appreciate additional features and advantages upon reading the following detailed description, and upon viewing the accompanying drawings.

**10** that is transmitting a communication signal via one or more antennas **12** to another wireless communication transceiver **14**, which receives the transmitted signal, plus other signals and noise, via one or more antennas **16**. While both of the transceivers **10** and **14** may be similarly configured, the transceiver **14** is shown in more detail, to highlight one or more embodiments of parametric G-RAKE processing as taught herein.

In simplified form, the transceiver **10** comprises one or more transmitter circuits **18**, one or more receiver circuits **20**, and additional control/processing circuits **22**, as needed. The transceiver **14**, while still simplified for clarity, illustrates more detail as a basis for discussion and comprises one or more transmitter circuits **24** and additional processing/control circuits **26** as needed, and more particularly comprises one or more receiver circuits **30** that include a radio processor circuit **32**, a parametric G-RAKE receiver circuit **34**, and additional receiver processing circuits **36** as needed, such as demodulation/decoding circuits that are configured to extract transmitted information bits from the traffic symbol estimates.

It should be understood that, for example, the transceiver **10** may comprise a radio base station in a wireless communication network and the transceiver **14** may comprise a mobile station—e.g., cellular radiotelephone, pager, palmtop computer, PDA, or other mobile communication device. More particularly, the transceiver **10** may comprise a base station configured according to the applicable WCDMA standards and the transceiver **14** may comprise a mobile station configured according to the applicable WCDMA standards. More generally, the transceiver **14** may comprise essentially any type of wireless communication device.

In the WCDMA context, the received communication signal of interest may comprise a HS-DSCH signal (used in HSDPA). The HS-DSCH is a shared packet data channel used to serve multiple users in time-multiplexed fashion based on scheduling which transmission intervals of the HS-DSCH are used to transmit data to particular ones of the users. Each transmission interval, which comprises three transmission slots, thus may be used to serve a different scheduled user. Of course, depending on the amount of data to be sent to a given user, multiple transmission intervals may be scheduled for the same user. For other examples of shared packet channels, a transmission interval may comprise a different number of transmission slots.

In other embodiments, the transceiver **14** may comprise a radio base station, e.g., a WCDMA radio base station. Application of the parametric G-RAKE receiver circuit **34** taught herein to base stations offers a number of advantages, especially under certain reverse link conditions. In particular, the receiver circuit **34** is advantageous when the physical quantities affecting the G-RAKE scaling parameters change rapidly between reverse link transmissions. The Enhanced Uplink (EUL) extension to the WCDMA standard represents one example where the signal conditions influencing the estimation of the scaling parameters can change significantly between transmission intervals of the reverse link signal(s) received at the base station.

Broadly, then,

As a starting point for understanding operations of the parametric G-RAKE receiver circuit **34**, it should be understood as comprising one or more processing circuits configured to estimate the scaling parameters on a per transmission interval basis, to prevent the scaling parameters estimated in one transmission interval of a received communication signal from being influenced by the scaling parameters estimated in another transmission interval of the received communication signal.

The term “processing circuit” should be broadly construed as connoting functional, but not necessarily physical, structure. For example, the parametric G-RAKE receiver circuit **34** may comprise one or more integrated circuit devices that are configured to carry out the signal processing taught herein. As such, the processing circuit(s) comprising the parametric G-RAKE receiver circuit **34** may comprise one or more microprocessors or DSPs, along with corresponding stored computer program instructions that configure the processing circuit(s) to carry out the desired signal processing. Alternatively, or additionally, the parametric G-RAKE receiver circuit **34** may be implemented using Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs), Complex Programmable Logic Devices (CPLDs), or the like.

Turning to processing-related details, it should be noted that the total impairment correlation taken across the RAKE fingers used for despreading the traffic channel component of the received communication signal may be denoted as an impairment correlation matrix R, where the total impairment equals interference plus noise. The impairment correlation matrix R can be expressed analytically as

*R=E* _{c}R_{ISI} *+N* _{0} *R* _{n } Eq. 1

where E_{c }is the total energy per chip period of the signal, N_{0 }is the noise power, and R_{ISI }and R_{n }are the modeled interference and noise correlation matrices, respectively. The interference matrix R_{ISI }depends on the pulse shape function, the chip period, the channel coefficients and delays. The noise matrix R_{n }depends on the pulse shape function and the delays. Note, assuming zero mean, the impairment correlation matrices are covariance matrices.

Because the matrix elements in the modeled impairment correlation matrices R_{ISI }and R_{n }can be obtained from quantities that are known or that can be learned (such as from the pilot channel component of the received signal), the estimation of R may be thought of the problem of estimating E_{c }and N_{0}. As a practical matter, the parametric G-RAKE receiver circuit **34** may use the pilot channel to make short-term estimates of the total impairment correlations R. These measured impairment correlations, which are short-term “snapshots,” may be denoted as {circumflex over (R)}, and parametric scaling factors α and β may be used in rewriting Eq. 1 to account for any additional scaling needed due to pilot channel estimation. With the parametric scaling factors, Eq. 1 may be rewritten as

*R=αR* _{ISI} *+βR* _{n } Eq. 2

where α is a first parametric G-RAKE scaling parameter used to scale the modeled interference correlations captured in R_{ISI}, and β is a second parametric G-RAKE scaling parameter used to scale the modeled noise correlations captured in R_{n}.

In Eq. 2, R_{ISI }is given by

and R_{n }is given by

*R* _{n }(*d* _{1} *,{tilde over (d)}* _{2})=*R* _{p}(*d* _{1} *−{tilde over (d)}* _{2})δ(α(*d* _{1})−α(*{tilde over (d)}* _{2})) Eq. 4

where the tilde indicates that {tilde over (d)}_{2 }may correspond to a different one of the receive antennas **16** than d_{1}, and where

represents the complex (pilot) channel medium model, R_{p }(τ) is the (filter) pulse shape autocorrelation function, T_{c }is the chip period of the received signal, d_{k }is the delay of the k^{th }G-RAKE finger, and α(d) indicates which one of the receive antennas **16** d corresponds to (e.g., receive antenna **1**, **2**, etc.). Also, note that the g values are channel coefficients corresponding to the pilot channel (i.e. channel coefficients scaled by pilot channel strength for the corresponding path delays).

**34** comprises a traffic correlation circuit **40**, a combiner circuit **42**, a finger placement circuit **44**, and G-RAKE processing circuits **46**.

The traffic correlation circuit **40** includes a number of traffic channel “fingers” (despreading correlators). Each traffic channel finger outputs despread traffic values γ from the received communication signal, r, output by the radio processor circuit **32**—e.g., n bit digital values corresponding to time-domain samples of the composite, antenna-received signal(s). Thus, for each channelization code of interest, the traffic correlation circuit **40** outputs a vector y of despread traffic values, with the length of the vector set by the number of fingers assigned to that channelization code.

In turn, the despread traffic values corresponding to the same symbol period for the same channelization code are combined by the combiner circuit **42** using interference-reducing combining weights. More specifically, the combiner circuit **42** uses combining weights that are formed based at least in part on parametrically determined received signal impairment correlations. The combining weights are given by w=R^{−1}h, where h is a vector of net channel coefficients corresponding to the G-RAKE finger delays. These combining weights are used to combine the despread traffic value vector elements to form traffic symbol estimates, z, which are provided to the additional receiver circuits **36**, e.g., for demodulation/decoding (z=w^{H}y).

The G-RAKE processing circuits **46** determine the appropriate combining weights based on channel estimation information and finger placement information from the finger placement circuit **44**, and on having access to samples of the received communication signal r. **46**, wherein it comprises a pilot correlation circuit **50**, a channel tracker **52**, a measured impairment correlation calculator **54**, a structured element calculator **56**, a scaling parameter calculator **58**, a modeled impairment correlation calculator **60**, a combining weight calculator **62**, and a signal quality estimator **64**, which may be a Signal-to-Interference Ratio (SIR) estimator.

In operation, samples of the received communication signal are provided to the pilot correlation circuit **50**, which correlates the received samples to a pilot or other reference signal spreading sequence, and removes symbol modulations to produce (pilot) correlation values. Correlations are performed at the finger delays provided by the finger placement circuit **44**. The channel tracker circuit **52** receives these pilot correlations and uses them to estimate or otherwise track channel coefficients for the received signal. These coefficients are provided to the measured impairment correlation calculator **54**, which also receives the pilot correlations. The calculator **54** is configured to obtain short-term measurements of impairment correlations for the received communication signal—i.e., “snapshot” impairment correlation measurements, {circumflex over (R)}, on a per transmission slot basis, for example—by subtracting channel estimates from the pilot correlations to obtain impairment samples and correlating the impairment samples with each other and with themselves.

Further, the structured element calculator **56** receives the channel estimates as well as finger delays from the finger placement circuit **44**, and it uses these to construct the elements corresponding to the modeled impairment correlation matrices R_{ISI }and R_{n}. The short-term impairment correlation measurements {circumflex over (R)} and the modeled impairment correlations R_{ISI }and R_{n }are provided to the scaling parameter calculation circuit **58**, which uses them to calculate the parametric scaling parameters needed, e.g., α to scale the R_{ISI }matrix and β to scale the R_{n}.

In turn, the modeled impairment calculator **60** is configured to generate a modeled estimate of impairment correlations for the received communication signal based on the scaling parameters and the modeled impairment correlations, e.g., R=αR_{ISI}+βR_{n}. From that calculation, the weight calculator **62** generates combining weights for combining the vector of despread traffic values output by the traffic correlation circuit **40**. Note, too, that channel estimates and the modeled estimate of impairment correlations also may be provided to the SIR calculator **64**, which produces a SIR estimate for power control purposes, among other uses.

With the above framework in mind,

In the illustration, one transmission interval comprises three transmission slots, and this illustrated configuration corresponds to the WCDMA standard adoption of three slots per Transmission Time Interval (TTI) for HSDPA transmissions. More particularly, HSDPA transmissions comprise three-slot TTIs of 2 ms width. Each TTI may be used to transmit data to a different scheduled user. Further, scheduled transmissions may use two or more consecutive TTIs, depending upon the amount of data to be transmitted to a particular user. Similar scheduled transmissions may be used on WCDMA reverse links, at least with the EUL provisions of Release 6 of the WCDMA standards.

In such circumstances—e.g., short TTIs, and rapidly changing transmission powers—carrying the measurements and/or calculations used to calculate the scaling parameters across transmission intervals can be problematic. Thus, the scaling parameter calculations taught herein provide a method for computing the parameters α and β in a manner that ensures good receiver performance for short TTIs and (potentially) rapid transmitter power variations.

More generally, the methods taught herein provide a basis for calculating robust (reliable) scaling parameters on a per transmission interval basis to prevent the scaling parameters calculated in one transmission interval from influencing their calculation in another transmission interval. In broad terms, the transmission interval/slot structure of the received signal may be imposed by the transmission method, or may be logically imposed by the parametric G-RAKE receiver circuit **34**.

Regardless, **100**), and then bounding or otherwise limiting the instantaneous value(s) (Step **102**). In one specific embodiment of this approach, each transmission interval comprises one or more transmission slots, and estimating the scaling parameters on a per transmission interval basis comprises estimating the scaling parameters on a per slot basis.

Estimating the scaling parameters on a per slot basis comprises obtaining measured impairment correlations for the transmission slot, generating instantaneous values for the scaling parameters based on the measured impairment correlations, and constraining the scaling parameters according to one or more defined limits. More particularly, the parametric G-RAKE receiver circuit **34** in **58** in _{ISI}) scaled by a first scaling parameter (α) and the modeled noise correlations (R_{n}) scaled by a second scaling parameter (β), and to solve the function for the first and second scaling parameters.

That is, given the receiver's ability to know or determine {circumflex over (R)}, R_{ISI}, and R_{n}, which includes its knowledge or determination of the channel (coefficients g_{1 }and delays τ_{1}), and the G-Rake finger delays d, every quantity in

*{circumflex over (R)}≈αR* _{ISI} *+βR* _{n } Eq. 5

is known except for α and β. To isolate the unknowns, the parametric G-RAKE receiver circuit **34** (e.g., the scaling parameter calculator **58**) can be configured to stack the columns of {circumflex over (R)}, R_{ISI}, and R_{n }as

where {circumflex over (r)}_{i }equals the i^{th }column of the measured impairment correlation matrix {circumflex over (R)}, r_{ISI,i }equals the i^{th }column of the modeled interference correlation matrix R_{ISI}, and r_{n,i }equals the i^{th }column of the modeled noise correlation matrix R_{n}. This system of equations can be solved via least squares (LS) fitting process. The LS fitting process can be expressed as

Note that this least square fitting need not be applied to all elements in the measured impairment correlation matrix, {circumflex over (R)}. For example, only the diagonal elements and the first off-diagonal elements could be used to perform the fitting operation. In general, {circumflex over (R)} is Hermitian symmetric, so only the diagonal and one of the upper or lower triangles would be used. This property can be used throughout, so that only unique matrix elements need to be computed and stored.

There are several aspects to the above-illustrated LS fitting process worth mentioning in further detail. First, the impairment covariance matrix R is of dimension N×N, where N is the number of probing fingers used by the parametric G-RAKE receiver circuit **34**. Simulation work done for parametric G-RAKE processing indicates that the instantaneous estimates of α and β provide good demodulation performance as long as N≧8. With the use of eight or more probing fingers, the parametric G-RAKE receiver circuit **34** may employ the fitting procedure described by Eq. 7 and Eq. 8 on a short-term basis (e.g., every slot, every other slot, etc.).

However, there are two aspects of the procedure described by Eq. 7 and Eq. 8 that bear additional explanation. First, the equations are given for simplicity of description only. In actuality, the parametric G-RAKE receiver circuit **34** would only use the matrix elements that corresponded to the upper (lower) triangular portion of the measured and modeled impairment correlation matrices. All of the upper (lower) triangular elements of these matrices generally should be used to ensure good demodulation performance.

A second point worth noting is that the parametric G-RAKE receiver circuit **34** has some a priori knowledge that it can exploit for simplification of its calculations, i.e., it may exploit the knowledge that the scaling parameters α and β are real-valued numbers to simplify the LS fitting problem. To illustrate this point, one may rewrite Eq. 5 as

*Re{R}+jIm{R}=α(Re{R* _{ISI} *}+jIm{R* _{ISI}})+β(*Re{R* _{n} *}+jIm{R* _{n}}) Eq. 9

where Re{ } is an operator that returns the real part of a complex value, Im{ } is an operator that returns the imaginary part of a complex value, and j is the square root of −1.

By collecting real and imaginary terms, one can reformulate Eq. 6 as

Because all the elements of Eq. 10 are real, the solution for α and β becomes

Note that the diagonal elements of {circumflex over (R)}, R_{ISI}, and R_{n }yield purely real equations.

With the above processing steps, as summarized in **34** allows for the measurements and calculations supporting the LS-based estimation of instantaneous values for the scaling parameters to be done on a per slot basis. In turn, the per slot calculations allow independence between the scaling parameters calculated in one transmission interval from those calculated in another transmission interval.

However, to ensure that the instantaneous values are robust, the method includes the illustrated bounding step. That is, the illustrated method constrains the scaling parameters according to one or more defined limits by limiting or otherwise bounding one or both of the instantaneous values. For example, the instantaneous value calculated for the second scaling parameter β may be constrained according to the value of the first scaling parameter α.

The step of thresholding (bounding) one or both of the scaling parameters may be optional for single-antenna implementations of the transceiver **14**. In such cases, the LS fitting step yielding the instantaneous estimates of α and β may provide sufficient demodulation performance. However, simulations for embodiments using multiple receive antennas **16** suggest that the block error rate (BLER) curves may have undesirably high error floors when the instantaneous values of α and/or β are used without bounding. The rise in error floor relates to variations in the ratio of the estimated scaling parameters,

It appears that the least squares estimation procedure can cause β to become quite small (with respect to α) or even negative. The phenomenon may be more pronounced in EUL and HS-DSCH scenarios, because the interference term is the dominant impairment in those cases. In any case, variations in the parameter ratio can cause large variations in soft information used by decoder circuits in the transceiver **14**. Such variations thus cause potentially poor decoder performance, resulting in high decoding error floors.

One example of parameter constraining is based on monitoring the relative values of α and β, and determining whether the ratio exceeds a certain (defined) thresholding value ρ. The thresholding value may be determined empirically, such as through simulation, for example. With this approach to parameter constraining, the instantaneous value of α as generated by the LS fitting process for a given slot or interval is kept the same (as long as it is positive), but the instantaneous value of β is set to

Here, it should be noted that, in a practical implementation, the parametric G-RAKE receiver circuit **34** uses the pilot channel in the estimation of α and β, and this practical detail means that α no longer represents the chip energy E_{c}, but rather the ratio of chip to pilot energy given as

Because the threshold(s) used to bound or otherwise limit β are predicated on

the parametric G-RAKE receiver circuit **34** must estimate E_{p}. A simple approach to estimating E_{p }is by averaging h^{H}h over time. This average can be computed with an exponential filter of the form

*E* _{p}(*n*)=γ*E* _{p}(*n−*1)+(1−γ)*h* ^{H} *h * Eq. 13

where γ is a constant between 0 and 1.

Incorporating the estimate of E_{p }into this embodiment gives the following relationship for β

In Eq. 14, the max (A,B) function returns the maximum (signed) value of A and B.

Processing begins with generating instantaneous estimates of α and β, such as on a per slot basis, according to the above LS fitting process (Step **110**), and continues with an estimation of pilot energy, E_{p}, such as detailed above (Step **112**). Processing continues with evaluation of the expression

where ρ is a desired (fixed or adjustable) thresholding constant (Step **114**). If the scaling parameter β is greater than the evaluated quantity, then the working value of β used for subsequent calculations is set to the instantaneous value of β (Step **116**). If not, then working value β is set to the constrained value

(Step **118**).

Of course, the processing logic of

In another embodiment, the parametric G-RAKE receiver circuit **34** monitors the ratio Ψ and β, and thresholds β. For example, if the ratio Ψ exceeds a certain threshold value ρ, or if the value of β<ε, β is set to

In another embodiment, the parametric G-RAKE receiver circuit **34** monitors the ratio Ψ. If the ratio Ψ exceeds the threshold value ρ, then the instantaneous values of α and β as calculated from the LS fitting process for the current transmission slot are replaced by the instantaneous values calculated in the previous slot. Of course, this embodiment generally is not used for the first slot of a given transmission interval to prevent scaling parameter values from being reused across transmission intervals. However, with that caveat, it is clear that previously calculated in-bounds scaling parameter values from a prior transmission slot of the same transmission interval can be re-used to replace out-of-bounds scaling parameter values calculated in a subsequent transmission slot.

In another embodiment, the parametric G-RAKE receiver circuit **34** computes the ratio Ψ for the first slot of a given transmission interval, e.g., a given TTI. If the ratio is less than a threshold value ρ, then the values α and β are frozen and used for all slots in the transmission interval. This embodiment thus may save computational cycles since the LS estimation of α and β need only be done once per transmission interval. However, values of α and β generally are not frozen from a prior transmission interval for carryover into a subsequent transmission interval. One exception to this is where the parametric G-RAKE receiver circuit **34** is processing a scheduled transmission (directed to it) over successive transmission intervals. Such circumstances might arise, for example, if the transceiver **14** is one of several scheduled users sharing a HS-DSCH in a WCDMA network. Also, it may be desirable as a general proposition, regardless of thresholding, to freeze α and β for an entire transmission interval and/or to limit the variation of Ψ.

In another embodiment, the parametric G-RAKE receiver circuit **34** monitors the ratio Ψ. If the ratio exceeds a threshold ρ, then the LS fitting procedure is reapplied to only estimate the parameter α. For the re-estimation of α, Eq. 2 becomes

Similar to the LS procedure described earlier for the estimation of α and β, the LS solution for α is given by

α=(*A* ^{T} *A*)^{−1} *A* ^{T} *p * Eq. 17

where

with k=1, . . . , K. Note that now A is a matrix with only one column and there is no matrix inversion involved because the quantity to invert is scalar.

In another embodiment, which is a variation of the above method, the parametric G-RAKE receiver circuit **34** imposes the condition that β is proportional to a according to some defined relationship. On that basis, the LS fitting process is modified by expressing the measured impairment correlations as a function of α only, and then solving for α using Eq. 16-Eq. 19. Once α is determined, then the defined relationship is used to calculate β.

In the above embodiments, robust estimates for the scaling parameters are obtained on a per slot basis, for example, by generating instantaneous values for α and β and applying bounding to one or both of the instantaneous values. Such bounding prevents, for example, the scaling parameter β from becoming too small in relation to α, and thus ensures that “good” working values of α and β are available for impairment correlation estimation and subsequent combining weight generation in each transmission slot.

In one or more further embodiments, robust estimates for the scaling parameters are obtained by accumulating data/measurements over two or more transmission slots within a given transmission interval. That is, rather than using a single slot to estimate α and β, the parametric G-RAKE receiver circuit **34** may be configured to use information from multiple slots to perform the LS fitting step(s) employed to determine estimates for the scaling parameters α and β within a given transmission interval.

**104**), such that the size of the sample sets used for estimation of the scaling parameters is larger. As an example, for a three-slot transmission interval, the parametric G-RAKE receiver circuit **34** can be configured to obtain measured impairment correlations from despread pilot (or traffic) values received over all three slots, thereby providing a larger number of samples for generating the snapshot impairment correlation measurements represented by the matrix {circumflex over (R)}.

The parametric G-RAKE receiver circuit **34** uses the accumulated measurements to estimate instantaneous values for the scaling parameters α and β (Step **106**), and, optionally, may apply bounding/limiting to one or both of the instantaneous values (Step **108**). However, using a larger sample set for estimating the instantaneous values of α and β, i.e., using more than one slot's worth of measurements, tends to lower the estimation “noise” of the scaling parameters. With the reduction in estimation noise, the LS fitting process used to obtain the scaling parameters generally yields inherently more robust results. Bounding thus may not be needed.

Mathematically, the parametric G-RAKE receiver circuit **34** solves the multiple-slot estimation problem given as

and A^{j }and p^{j }represent the A matrix and p vector for the j^{th }slot of the given transmission interval.

As a variation of this method, the parametric G-RAKE receiver circuit **34** may be configured to obtain LS estimates of α and β for each slot in a transmission interval (e.g. B=A^{j}, x=p^{j}), and then average the estimates together. In mathematical form, this variant is given by

where k represents the current slot index, and j indexes from the beginning of the transmission interval—e.g., TTI—to the current slot index. Note, too, that averaging over multiple slots still represents a per transmission interval solution, as slot-to-slot average generally is not carried over across transmission intervals.

In still other embodiments, the parametric G-RAKE receiver circuit **34** can be advantageously configured for use with multiple receive antennas **16**. Consider two receive antennas **16**, for example, where Eq. 2 becomes

The elements of the off-diagonal blocks of the correlation matrix R_{n }correspond to the noise correlation between fingers of different antennas, and are equal to zero. All previous embodiments may be used with multiple receive antennas. Another embodiment considers setting to zero the off-diagonal elements of the measured impairment correlation matrix R as follows

Then the estimation of the scaling parameters α and β may be done using the LS fitting process described earlier herein.

It should be noted that this method biases the estimation of the parameter α towards lower values, but it works well in that it reduces variation in the ratio of the scaling parameters (with respect to the case that all elements of the R matrix are used in the LS estimation). If the LS fitting process yields negative values for the scaling parameters, the parametric G-RAKE receiver circuit **34** may be configured simply to set them to zero.

As a further note, the method immediately above represents a one-step process for obtaining robust estimates of the scaling parameters α and β. However, it will be appreciated that a constraining step may be performed as an optional, second step, wherein the value of α and/or β is bounded or otherwise limited to ensure good receiver performance.

In another embodiment, the parametric G-RAKE receiver circuit **34** uses an iterative procedure to estimate the scaling parameters. Consider a case where the transceiver **14** uses two receive antennas **16**. The elements of the off-diagonal blocks of the correlation matrix R_{n }correspond to the noise correlation between fingers of different antennas. These values are equal to zero. Therefore, for the two-antenna case, Eq. 5 becomes Eq. 24, which is repeated here as

One sees from the above expression that the value of β does not depend on the off-diagonal blocks. Thus, estimation of β can be done using only the diagonal blocks. Two steps are used to estimate α and β. First, only the diagonal blocks are used to estimate both parameters, with these initial estimates denoted as α_{1 }and β_{1}. A suitable expression for this is given as

Then, β is set to the estimated value β=β_{1 }and Eq. 26 is used to re-estimate the parameter α, denoted as α_{2}. The scaling parameter β is then re-estimated by setting α=α_{2 }and using Eq. 27 to determine β_{2}. The iteration between fixing β and re-estimating α using Eq. 26, then fixing α and re-estimating β using Eq. 27 can be repeated a pre-determined number of times or until satisfying a defined stopping criteria. In general, then, estimating the scaling parameters on a per transmission interval basis may be based on initially estimating first and second scaling parameters and then iteratively using the estimate for one of the first and second scaling parameters to revise the estimate for the other one of the first and second scaling parameters

_{1 }and β_{1 }from the diagonal blocks as described above (Step **120**). Processing continues with the calculation of the ratio

(Step **122**). The value β_{2}=β_{1 }is then used to estimate α_{2 }via the LS fitting process (Step **124**), and a new ratio

s calculated (Step **126**). The two ratios Ψ_{1 }and Ψ_{2 }are tested in comparison to a defined threshold ρ (Step **128**). If both of the ratios exceed the value of ρ, then the scaling parameters α and β corresponding to the smallest one of the two ratios are used, with β constrained according to a defined threshold (bounded) as a function of α (Step **130**).

Alternatively, if only one of the ratios does not exceed the threshold ρ, then the values of the scaling parameters α and β corresponding to that ratio are used. Thus, if Ψ_{1 }exceeds the threshold and Ψ_{2 }is below the threshold (checked by the combination of Steps **128** and **132**), then the scaling parameters corresponding to Ψ_{2 }are used (Step **134**). Conversely, if Ψ_{2 }exceeds the threshold and Ψ_{1 }is below the threshold (checked by the combination of Steps **128** and **132**), then the scaling parameters corresponding to Ψ_{1 }are used (Step **136**).

In another iterative solution embodiment, the parametric G-RAKE receiver circuit **34** is configured to improve its estimation of the scaling parameters α and β by exploiting additional information available to it. For example, the received communication signal generally includes error correction codes and/or error detection codes that may be used to improve scaling parameter estimation.

For example, assuming that a Cyclic Redundancy Check (CRC) value is provided for each (received signal) coding block, the parametric G-RAKE receiver circuit **34** can improve its scaling parameter estimation by trying different candidate pairs of the scaling parameters α and β. For each candidate pair, the received signal is demodulated and decoded, and the CRC is checked. All candidate pairs of parameters corresponding to a CRC that does not check are discarded. The process continues for all candidate pairs until the CRC checks or all candidate pairs are processed. Note that one coding block can correspond to multiple transmission slots and different transmission slots can use different scaling parameters. In such cases, the candidate pair refers to the set of pairs of parameters of all slots in a block.

There are several ways to decide on a good candidate set. For instance, assume that the parametric G-RAKE receiver circuit obtains initial estimates of α and β. With these initial estimates, it can then create a set of candidate pairs by modifying the nominal values α and β, such as by α=α+ε_{1 }and β=β+ε_{2}, where ε_{1 }and ε_{2 }can be chosen as fractions of α and β respectively. Another way is to try different ratios of α to β. Of course, other approaches to obtaining the candidate set of scaling parameter pairs may be used, and those skilled in the art will appreciate that the above example is non-limiting.

Yet another embodiment of scaling parameter estimation uses a two-step scaling parameter determination process that is valid for real-valued receive filters. (Real-valued receiver filters are reasonable assumption for most WCDMA receiver implementations, for example.) Recalling the impairment correlations expression given in Eq. 2, if the receive filter is real-valued, then R_{n }is real. Since β must be real, the imaginary entries of the impairment correlation matrix R are a function of α only. Conceptually, the parametric G-RAKE receiver circuit **34** can solve for α first using the imaginary entries, and then solve for β using the real entries plus the value computed for α. In terms of a detailed procedure, this embodiment consists of the following steps: (1) solve a least squares problem for α given the over-determined system of equations imaginary ({circumflex over (R)})≈α[imaginary (R_{ISI})]; and (2) using the α from Step (1), solve a least squares problem for β given the over-determined system of equations real ({circumflex over (R)}−αR_{ISI})≈βR_{n}.

Regardless of which one or combination of the above embodiments is implemented, the present invention teaches a wireless communication device or system, e.g., the transceiver **14**, that uses parametric G-RAKE processing based on per interval estimation of the scaling parameters. Per interval estimation improves the impairment correlation and corresponding combining weight calculations performed as part of G-RAKE processing in circumstances where the conditions bearing on scaling parameter estimation potentially change rapidly between transmission intervals. As such, scaling parameter estimation on a per interval basis generally prevents the scaling parameters estimated in one transmission interval of a received communication signal from being influenced by the scaling parameters estimated in another transmission interval of the received communication signal.

On that point, the term “per interval basis” as used herein should be given broad construction. For example, some embodiments of scaling parameter estimation taught herein do carry over scaling parameters from one time slot to another, i.e., the scaling parameter(s) estimated in one time slot are reused in one or more subsequent time slots. However, to the extent that new values for the scaling parameter(s) are being calculated within a given transmission interval, those calculations are driven by the signal conditions prevailing during all or part of that given transmission interval.

With that broad understanding in mind, those skilled in the art will appreciate that the present invention is not limited to the foregoing features and advantages. Instead, the present invention is limited only by the following claims and their legal equivalents.

Patent Citations

Cited Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|

US5572552 * | Jan 27, 1994 | Nov 5, 1996 | Ericsson Ge Mobile Communications Inc. | Method and system for demodulation of downlink CDMA signals |

US5809020 * | Mar 18, 1996 | Sep 15, 1998 | Motorola, Inc. | Method for adaptively adjusting weighting coefficients in a cDMA radio receiver |

US6006110 * | Oct 18, 1996 | Dec 21, 1999 | Cisco Technology, Inc. | Wireless communication network using time-varying vector channel equalization for adaptive spatial equalization |

US6101399 * | Jun 16, 1995 | Aug 8, 2000 | The Board Of Trustees Of The Leland Stanford Jr. University | Adaptive beam forming for transmitter operation in a wireless communication system |

US6363104 * | Oct 2, 1998 | Mar 26, 2002 | Ericsson Inc. | Method and apparatus for interference cancellation in a rake receiver |

US6470044 * | Jan 15, 1999 | Oct 22, 2002 | Sharp Laboratories Of America, Inc. | Computationally parsimonious forward link receiver for DS-CDMA systems and method for same |

US6529545 * | Mar 30, 2001 | Mar 4, 2003 | Nokia Networks Oy | Rake receiver |

US6683924 * | Oct 19, 1999 | Jan 27, 2004 | Ericsson Inc. | Apparatus and methods for selective correlation timing in rake receivers |

US6714585 * | Jun 25, 1999 | Mar 30, 2004 | Ericsson Inc. | Rake combining methods and apparatus using weighting factors derived from knowledge of spreading spectrum signal characteristics |

US7010019 * | Sep 18, 2002 | Mar 7, 2006 | Telefonaktiebolaget Lm Ericsson (Publ) | Assessment of delay estimation quality using interference estimates |

US7046963 * | Jan 29, 2003 | May 16, 2006 | Texas Instruments Incorporated | Method and apparatus of signal estimation over generalized fading channel |

US7349379 * | Aug 13, 2003 | Mar 25, 2008 | Texas Instruments Incorporated | Channel normalization |

US7751511 * | Oct 19, 2006 | Jul 6, 2010 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for selecting a subset of modeled impairment correlation terms for use in received signal processing |

US20010028677 * | Apr 30, 2001 | Oct 11, 2001 | Wang Yi-Pin Eric | Apparatus and methods for finger delay selection in RAKE receivers |

US20020057735 * | Dec 3, 2001 | May 16, 2002 | Olli Piirainen | Optimizing of channel equalizer |

US20020080863 * | Dec 14, 2001 | Jun 27, 2002 | Nielsen Jorgen S. | Adaptive generalized matched filter rake receiver system and method |

US20020122470 * | Dec 29, 2000 | Sep 5, 2002 | Heikkila Markku J. | Method and apparatus for providing blind adaptive estimation and reception |

US20020150176 * | Feb 27, 2002 | Oct 17, 2002 | Thales | Method and device for the estimation of a propagation channel from its statistics |

US20030016740 * | May 4, 2001 | Jan 23, 2003 | Jeske Daniel R. | Method of estimating a signal-to-interferenceratio (SINR) |

US20030035468 * | May 17, 2001 | Feb 20, 2003 | Corbaton Ivan Jesus Fernandez | System and method for adjusting combiner weights using an adaptive algorithm in wireless communications system |

US20050101279 * | Sep 16, 2003 | May 12, 2005 | Lee Kuo H. | Receiving method and receiver |

US20050201447 * | Mar 12, 2004 | Sep 15, 2005 | Cairns Douglas A. | Method and apparatus for parameter estimation in a generalized rake receiver |

US20060007990 * | Sep 2, 2005 | Jan 12, 2006 | Telefonaktiebolaget Lm | Method and apparatus for scaling parameter estimation in parametric generalized rake receivers |

US20060182204 * | Apr 12, 2006 | Aug 17, 2006 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for selecting demodulation processing delays in a receiver |

Referenced by

Citing Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|

US7539240 | Mar 12, 2004 | May 26, 2009 | Telefonaftiebolaget Lm Ericsson (Publ) | Method and apparatus for parameter estimation in a generalized rake receiver |

US7688881 | Jun 30, 2006 | Mar 30, 2010 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for interference estimation in a generalized RAKE receiver |

US7738534 | Dec 5, 2006 | Jun 15, 2010 | Telefonaktiebolaget Lm Ericsson (Publ) | Multi-transmitter interference suppression using code-specific combining |

US7751463 | Dec 5, 2006 | Jul 6, 2010 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for suppressing interference based on channelization code power estimation with bias removal |

US7751511 | Oct 19, 2006 | Jul 6, 2010 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for selecting a subset of modeled impairment correlation terms for use in received signal processing |

US7848387 * | May 21, 2007 | Dec 7, 2010 | Telefonaktiebolaget Lm Ericsson (Publ) | Receiver parametric covariance estimation for transmit diversity |

US7848389 | Sep 2, 2005 | Dec 7, 2010 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for scaling parameter estimation in parametric generalized rake receivers |

US7949063 * | Feb 27, 2007 | May 24, 2011 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for impairment correlation estimation in MIMO systems |

US7983208 | Nov 6, 2007 | Jul 19, 2011 | Telefonaktiebolaget Lm Ericsson (Publ) | MMSE channel estimation in a communications receiver |

US7995641 | Nov 6, 2007 | Aug 9, 2011 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for code power parameter estimation for received signal processing |

US8009722 | Dec 12, 2007 | Aug 30, 2011 | Telefonaktiebolaget Lm Ericsson (Publ) | Multi-pass parameter estimation for G-Rake receiver |

US8040938 | Dec 12, 2006 | Oct 18, 2011 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for extended least squares estimation for generalized rake receiver parameters using multiple base stations |

US8045600 * | Apr 29, 2008 | Oct 25, 2011 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for compensating for processing timing misalignment in a communication receiver |

US8098715 | Jun 9, 2008 | Jan 17, 2012 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for estimating impairment covariance matrices using unoccupied spreading codes |

US8229044 | Jun 14, 2007 | Jul 24, 2012 | Telefonaktiebolaget Lm Ericsson (Publ) | Efficient covariance computation by table lookup |

US8275023 * | Dec 12, 2006 | Sep 25, 2012 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for shared parameter estimation in a generalized rake receiver |

US8428106 | Jun 13, 2008 | Apr 23, 2013 | Telefonaktiebolaget L M Ericsson (Publ) | Efficient method for forming and sharing impairment covariance matrix |

US8483641 * | Jul 19, 2011 | Jul 9, 2013 | Marvell International Ltd. | Validation and stabilization of noise matrices |

US8526552 | Jul 7, 2010 | Sep 3, 2013 | Marvell International Ltd. | Noise estimation in communication receivers |

US8750440 | Jul 30, 2013 | Jun 10, 2014 | Marvell International Ltd | Noise estimation in communication receivers |

US8855172 | Dec 9, 2011 | Oct 7, 2014 | Telefonaktiebolaget Lm Ericsson (Publ) | Non-redundant equalization |

US8964912 | Sep 2, 2005 | Feb 24, 2015 | Telefonaktiebolaget Lm Ericsson (Publ) | Adaptive timing recovery via generalized RAKE reception |

US8995552 | Jan 25, 2013 | Mar 31, 2015 | Telefonaktiebolaget L M Ericsson (Publ) | Method and apparatus for impairment correlation estimation in MIMO systems |

US9077422 | Jun 11, 2013 | Jul 7, 2015 | Marvell International Ltd | Validation and stabilization of noise matrices |

US20050201447 * | Mar 12, 2004 | Sep 15, 2005 | Cairns Douglas A. | Method and apparatus for parameter estimation in a generalized rake receiver |

US20060007990 * | Sep 2, 2005 | Jan 12, 2006 | Telefonaktiebolaget Lm | Method and apparatus for scaling parameter estimation in parametric generalized rake receivers |

US20070189364 * | Dec 12, 2006 | Aug 16, 2007 | Wang Yi-Pin E | Method and Apparatus for Shared Parameter Estimation in a Generalized Rake Receiver |

US20100027590 * | Nov 7, 2006 | Feb 4, 2010 | Girum Alebachew | Modified sir values for fast power control |

US20120238227 * | Dec 21, 2009 | Sep 20, 2012 | Thomson Licensing | Autocorrelation-based spectrum sensing for fm signals |

EP2062368A2 * | Jul 17, 2007 | May 27, 2009 | Telefonaktiebolaget LM Ericsson (PUBL) | Method and apparatus for shared parameter estimation in a generalized rake receiver |

EP2074702A1 * | Sep 28, 2007 | Jul 1, 2009 | Telefonaktiebolaget LM Ericsson (PUBL) | Doppler frequency control of g-rake receiver |

EP2327178A1 * | Sep 16, 2009 | Jun 1, 2011 | Telefonaktiebolaget L M Ericsson (PUBL) | Interference handling using a priori knowledge on interfering signal characteristics |

WO2008031761A1 * | Sep 7, 2007 | Mar 20, 2008 | Ericsson Telefon Ab L M | Method and apparatus for extended least squares estimation for generalized rake receiver parameters using multiple base stations |

WO2008044989A1 | Sep 28, 2007 | Apr 17, 2008 | Ericsson Telefon Ab L M | Doppler frequency control of g-rake receiver |

WO2008046910A1 * | Oct 19, 2007 | Apr 24, 2008 | Ericsson Telefon Ab L M | Method and apparatus for selecting a subset of modeled impairment correlation terms for use in received signal processing |

WO2008057018A1 * | Nov 7, 2006 | May 15, 2008 | Ericsson Telefon Ab L M | Modified sir values for fast power control |

WO2008151983A1 * | Jun 5, 2008 | Dec 18, 2008 | Ericsson Telefon Ab L M | Efficient covariance computation by table lookup |

WO2008154653A2 * | Jun 13, 2008 | Dec 18, 2008 | Ericsson Telefon Ab L M | Efficient method for forming and sharing impairment covariance matrix |

WO2009061263A1 * | Oct 20, 2008 | May 14, 2009 | Ericsson Telefon Ab L M | Method and apparatus for code power parameter estimation for received signal processing |

Classifications

U.S. Classification | 375/148, 375/150 |

International Classification | H04B1/707 |

Cooperative Classification | H04B1/712, H04B2201/709727, H04L25/0222, H04B1/7103, H04L25/0228, H04L25/0212 |

European Classification | H04L25/02C6, H04B1/7103, H04L25/02C3 |

Legal Events

Date | Code | Event | Description |
---|---|---|---|

Sep 2, 2005 | AS | Assignment | Owner name: TELEFONAKTIEBOLAGET LM ERICSSON (PUBL), SWEDEN Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:COZZO, CARMELA;CAIRNS, DOUGLAS A.;BOTTOMLEY, GREGORY E.;AND OTHERS;SIGNING DATES FROM 20050826 TO 20050829;REEL/FRAME:016960/0064 |

Jun 9, 2014 | FPAY | Fee payment | Year of fee payment: 4 |

Rotate